GitHub’s Post

Transcript

As AI developer tools become more capable and accessible, they enable more people to create software through natural language. But when the code is generated, the user has to do something with it, like building, running, and deploying it. That's why GitHub Next have built GitHub Spark an AI-native app platform to help anyone build use and share highly personalized microapps. Let's see how it works. From the homepage, we can see our existing Sparks (or microapps) and can describe what we want to create. So let's build an app that helps a user review the cities that they visited. Rather than being provided with the code, we're provided with an app that we can interact with immediately. Now the app is made up of custom components, providing a consistent feel. I can set the theme to light or dark mode. Or let it respect my operating system���s theme preference. But I even have access to a theme editor to control a few aspects like the primary accent colors or border radiuses. Now this looks like a good start, but let's iterate on the idea by capturing the latitude and longitude of a city. Now we already added some data, but it has no latitude or longitude. So let's click on the data tab and delete it. By the way, notice how it's storing state without us having to manage any databases or connection strings? Cool. So let's add some more reviews. Looks like the coordinates have been calculated, so let���s iterate a bit further. Let's allow the user to update and delete items in their cities list and plot those ratings onto a map. As we view the travel log, we can edit existing items and delete them as well. But let's go and add it back in. Now when we hit the toggle, we're taken to a map view and can see Barcelona and Rome plotted on the map. Now, I might want some suggestions on cities to explore. So let's ask Spark to help there too. and add some capabilities into our app. We'll ask it to suggest three cities based on my existing ratings and reviews. Now, as we generate city suggestions, notice the multicolored animation around the button. That shows we're using AI to generate a result. And it looks like there's some great destinations. Now let's do one more thing. Let's provide a vague request to make the app feel more sunny and fun. So let's tap on that double plane button and generate variants. This is useful as my request was ambiguous and our intent wasn't fully clear. This provides us with several alternate paths that we can explore When we're happy with an option, we can select it and continue building out the app. As you're iterating on your app, you'll notice a history of changes on the left hand side. As we move between them, we can explore the revision prompts and the resulting state that was generated. So if you want to revert some changes or explore an alternate path, you are empowered to do that. But this looks good to me. So let me rename it and let me make sure to star it so it's easily accessible from my dashboard. GitHub Spark has helped me create my family travel review app, immediately usable from my tablet without any deployment needs. Want to learn more? Check out more details by going to gh.io/spark